The efficiency of greedy best response algorithm in road traffic assignment

In this work, we investigate the problem of the integer road traffic assignment. So, we model the interaction among the road users sharing the same origin-destination pair, as a symmetric network congestion game. We focus on Rosenthal's results to guarantee the existence of a pure Nash equilibrium (PNE). Then, we study the behaviour of an algorithm based on greedy best response (GBR) in finding PNE. In previous studies, the efficiency of GBR to compute a PNE of a symmetric network congestion game in series-parallel networks is proved. In our work, another approach is used to demonstrate its efficiency in more general networks. It is shown that the non-series parallel networks can be classed into two types. The conditions that make GBR succeeds for each type is then drawn. The advantage of GBR-algorithm is that provides a better approximation of the equilibrate assignment, since it is integer.